ste

Upstream audit

This document records the audit of the related projects. The audit came before the first vocabulary file of this project.

Audit date: 2026-09-02. All the data comes from the GitHub API and from the files of each repository, at the commit that this document gives below.

Result

This project adopts no upstream code and no upstream data. A person wrote all the rule data in internal/checker/rules/ by hand for this project. Thus there is no THIRD_PARTY_NOTICES.md file. Make that file before the first import of an upstream data file.

Method

Record the URL, the commit, the language, the license, the vocabulary path, and the tests of each candidate.
Keep the license of the code separate from the provenance of the data. A permissive license on a repository is not proof that the dictionary data in that repository is safe to redistribute.
Give each candidate one of three results: reuse, borrow design only, or do not use.

Candidates

sourdough-bread/asd-ste100-checker

Item	Value
Commit	`e193ecdd66b0` (2026-07-24)
Language	Python
Code license	Apache-2.0
Vocabulary path	`ste100/` data modules, extracted from the specification
Result	do not use (data), borrow design only (interfaces)

The repository holds ASD-STE100-ISSUE-9.pdf at its root. This is the specification, which is under copyright. The repository says that it commits dictionary and rule data that comes from that PDF, and it gives a warning about the risk. The Apache-2.0 license applies to the code of the project. It does not give the project the right to sub-license the specification data.

Kept as ideas only: the rule identifiers, the structured result format, and the later MCP and LSP interfaces.

valeratrades/ste_checker

Item	Value
Commit	`3adb827991ee` (2026-08-21)
Language	Rust
Code license	Blue Oak Model License 1.0.0
Vocabulary path	`ste_checker/src/wordset.rs` with the openste word list
Result	borrow design only

Correction to the plan: the license is Blue Oak 1.0.0, not MIT. The MIT statement in the plan refers to the openste word list that this project vendors, not to the project itself.

Kept as ideas only: the glossary workflow, and the labeled test corpus with a corpus.truth file. This project uses the same idea in testdata/*.expected.json.

swarooppatilx/ste100

Item	Value
Commit	`6bbed1e63941` (2026-08-09)
Language	Python
Code license	MIT
Vocabulary path	`src/data/approved.json`, `unapproved.json`, `technical.json`
Result	borrow design only

The repository gives no statement about the source of its three JSON data files. The MIT license of the code is not enough proof for the data. This project does not use that data.

Kept as ideas only: one test file for each rule, and the annotated sample files.

stazelabs/ste

Item	Value
Commit	`922c9773d000` (2026-07-30)
Language	Go
Code license	MIT
Vocabulary path	none. The repository says that it contains no ASD-STE100 specification data.
Result	borrow design only

This is the nearest implementation reference: a Go CLI with no dependencies, byte spans, JSON output, and a score for each 100 words. This project copied no code. The score for each 100 words and the .ste.yml config file are the same ideas.

The README of that project is also one of the unlabeled text samples that found false positives. See docs/evaluation.md.

openste/openste

Item	Value
Commit	`19f01781a616` (2026-03-16)
Language	data only
Code license	MIT (`LICENSE`, “Copyright (c) 2026 openSTE.org”)
Vocabulary path	`vocabulary/openste.json`, 436 kB
Result	do not use in v1. Possible `reuse` after a provenance review.

Content of the file: set_name “openste - v1.01”, 1951 word records (909 approved, 1042 unapproved), each with a spacypos part of speech, and 1589 alternative-word records.

This is the strongest vocabulary lead, but two facts stop its use in v1:

The repository gives no statement about the source of the word list. The README and the vocabulary/README.md describe the idea of a controlled vocabulary, but they do not say who made the list or from what.
The CHANGELOG.md puts the “Controlled vocabulary list” in its Planned section, but vocabulary/openste.json already holds 1951 records with an approved/unapproved status. The structure is very near to the structure of the specification dictionary.

The MIT license is a statement by the openste project. It is not proof of the provenance of the data. Because v1 does not need a full dictionary rule, the safe decision is to wait.

Condition for a later reuse: the openste project states the source of its word list. As an alternative, the user imports the list with a command. In both cases, write THIRD_PARTY_NOTICES.md first.

TechScribe term checker (design page)

Item	Value
Source	https://www.simplified-english.co.uk/design.html
License	commercial and proprietary
Result	borrow design only

Kept as an idea only: a technical noun and a technical verb need different project configuration. docs/glossary.yml has the two lists allow.nouns and allow.verbs for this reason.

The specification itself

The owner of this repository has a licensed copy of ASD-STE100 Issue 9. On 2026-09-02, a review of that copy corrected the rule numbers of this tool. The review also found which rules a deterministic checker can check. The result is in rules.md and in evaluation.md.

What entered this repository from that work:

Rule numbers and the subject of each rule. A rule number is a fact about the specification. It is not the text of the specification.
Numeric limits, such as the 20-word and 25-word sentence limits, and the count rules of section 8.
Statements in our own words about what each rule requires.

What did not enter this repository:

The text of the specification, its tables, and its example sentences.
The dictionary of part 2. Part 2 gives about 2000 words in a form that a program can parse. It is the largest and most valuable part of the specification, and it stays out of this repository.

The local copy and its Markdown conversion are in .standard/, which .gitignore excludes. This repository is public, and git must never see that directory.

The terms of ASD, in detail

The copyright page of Issue 9 gives three statements that decide what this repository can hold:

The bar. “no reproduction or publication of it, in whole or in part, shall be made without the written authority of an officer” of ASD.
Distribution. ASD gives the specification free of charge to each writer and user through asd-ste100.org. But: “Unauthorized distribution of ASD-STE100, direct or through different websites or portals, is strictly prohibited without written permission from the STEMG.” A public repository is such a portal.

Special usage rights. ASD gives irrevocable permission to 8 groups. They can use, reproduce, or publish the document, in whole or in part, free of charge. The groups include the member companies of ASD, AIA, AIAC, and ICCAIA. They also include the customers of those companies, and universities and research institutes for educational purposes. The owner of this repository is not in one of those groups.

Statement 2 is the reason that this repository holds no dictionary. Two paths can change that: written permission from the STEMG, or an owner who is in one of the groups of statement 3.

Statement 2 also gives the design of the ste dict command. The specification is free for each user, and a user can make the index from their own copy. The tool ships the parser, the user supplies the data, and nothing moves from one user to another.

What a rule number is

A rule number, a numeric limit, and a statement in our own words are facts about the specification. Copyright protects the expression of an idea, and not the idea. The dictionary is different. It is a compilation, and ASD is in Brussels. Thus the database right of the European Union can protect the investment in it, and that right is separate from copyright. This is the reason for more care with the dictionary than with the rules.

This is not legal advice.

Dependencies

Module	Version	License	Function
`github.com/yuin/goldmark`	v1.8.5	MIT	Reads the Markdown
`github.com/goccy/go-yaml`	v1.19.2	MIT	Reads the glossary
`github.com/spf13/pflag`	v1.0.10	BSD-3-Clause	Reads the command flags
`github.com/adrg/xdg`	v0.5.3	MIT	Gives the data directory of the user

Each of the 4 is code, and not data. None of them has ASD-STE100 data. The tool used its own reader for Markdown and for YAML before version 0.4.0. A CommonMark parser is better, because 5 of the 8 defects of the first versions came from that hand-written reader.

gopkg.in/yaml.v3 is not in this list. That project became an archive in April 2025.

Data in this project

File	Source	Risk
`internal/checker/rules/verbdata.go`	Written by hand. English irregular participles, auxiliary verbs, and a list of 24 phrasal verbs.	Low. English grammar is not under copyright.
`internal/checker/rules/terms.go`	Written by hand. 27 words and 11 word groups that have a shorter replacement.	Low. It is not the ASD-STE100 dictionary and it does not come from it.
`internal/checker/rules/spelling.go`	Written by hand. 54 British spellings, 17 word groups that use a noun for an action, and 8 Latin abbreviations.	Low. The lists are examples of the rule, and not the dictionary.
`docs/glossary.yml`	Written by hand for this repository.	None.

This tool does not ship the ASD-STE100 approved-word dictionary. It cannot tell you if a word is in that dictionary.

This site is open source. Improve this page.